Karmiloff-Smith & Thomas, page 1
To appear in Special Issue of Development and Psychopathology:
“Contributions of Nature to Developmental Theory”
What can developmental disorders tell us about the neurocomputational
constraints that shape development? The case of Williams syndrome
Annette Karmiloff-Smith1 & Michael Thomas2
1
Neurocognitive Development Unit, Institute of Child Health, London
2
Birkbeck College, University of London
Karmiloff-Smith & Thomas, page 2
Abstract
The uneven cognitive phenotype in the adult outcome of Williams syndrome has led
some researchers to make strong claims about the modularity of the brain and the
purported genetically determined, innate specification of cognitive modules. Such
arguments have particularly been marshalled with respect to language. We challenge
this direct generalisation from adult phenotypic outcomes to genetic specification, and
consider instead how genetic disorders provide clues to the constraints on plasticity
that shape the outcome of development. We specifically examine behavioural studies,
brain imaging and computational modelling of language in Williams syndrome, but
contend that our theoretical arguments apply equally to other cognitive domains and
other developmental disorders. While acknowledging that selective deficits in normal
adult patients might justify claims about cognitive modularity, we question whether
similar, seemingly selective deficits found in genetic disorders can be used to argue
that such cognitive modules are pre-specified in infant brains. Cognitive modules are,
in our view, the outcome of development, not its starting point. We note that most
work on genetic disorders ignores one vital factor: the actual process of ontogenetic
development, and argue that it is vital to view genetic disorders as proceeding under
different neurocomputational constraints, not as demonstrations of static modularity.
Karmiloff-Smith & Thomas, page 3
I. Introduction
At times, genetic mutations give rise to atypical development from embryogenesis
onwards. Some are inherited but others are due to purely random events. One such
genetic disorder – Williams syndrome (WS) – is caused by the chance misalignment
during meiosis of identical flanking regions surrounding some 25 genes on one copy
of chromosome 7q11.23 (see Donnai & Karmiloff-Smith, 2000, for review). Because
of its uneven cognitive phenotype in the adult outcome, WS has given rise to strong
claims about the modularity of the human mind and the purported genetically
determined, innate specification of such modules (Pinker, 1994, 1997, 1999; Smith,
1999; Smith & Tsimpli, 1995). But can one generalise directly from adult phenotypic
outcomes to genetic specification in this way? In this paper we argue against the
simplicity of such claims and consider instead the way in which genetic disorders are
informative about the constraints on plasticity that shape the outcome of processes of
development. In doing so, we raise a number of crucial questions. First, can one use
the selective deficits found in the mature, previously normally developed brain of
adult neuropsychological patients to make claims about the cognitive modularity of
the human brain? Second, if similar, seemingly selective deficits are also found in
genetic disorders, can these be used to argue that such cognitive modules are
genetically determined and pre-specified in the infant brain? We will answer in the
affirmative to the first question, with a caveat about whether such pure deficits truly
exist. But our response will be negative to the second question, arguing that most
work on genetic disorders ignores one vital factor: the process of ontogenetic
development. In particular, we submit that it is vital to view genetic disorders as
proceeding developmentally under different neurocomputational constraints, not as
Karmiloff-Smith & Thomas, page 4
demonstrations of static modularity. We will illustrate our arguments with the case of
language acquisition in Williams syndrome, although we maintain that the general
tenets apply equally to other domains of cognition and to other developmental
disorders.
II.
Selective Deficits in Adult Neuropsychological Patients
It is now well established that adults who had previously developed normally can, in
the case of focal damage due to stroke or trauma, display selective behavioural
deficits (Cappelletti, Butterworth & Kopelman, 2001; Cipolotti, Butterworth &
Warrington, 1995; Coltheart, 2002; Patterson, 1981; Rapp & Caramazza, 2002;
Shallice, 1988). The existence of such patients has led to the postulation that the brain
is composed of modules, each dedicated to a specific kind of input processing.
Patients with severe agrammatism may present with other aspects of cognition,
including non-grammatical aspects of language, that seem to function normally.
Likewise, other patients may have serious word-finding difficulties but their syntactic
expression is fluent. This suggests that, by adulthood, specialised functions have
become relatively localised to specific brain regions (in this case, that processing of
grammar and word-specific knowledge relies on distinct underlying circuitry).
However, it is important to recall that such an inference from selective deficit to
normal structure is predicated on the assumption that impaired behaviour can be
traced to damaged underlying circuitry and that intact behaviour can be traced to
residual normally functioning circuitry. This assumption corresponds to the idea that
following focal damage, the rest of the system continues to function exactly as it did
before the brain insult, thereby giving rise to scores in the normal range. Of course, in
attempting to establish that apparently intact functions are indeed working just as they
Karmiloff-Smith & Thomas, page 5
did before, the researcher is always at the mercy of the sensitivity of the measurement
scale used. Our own work on adults with genetic disorders has demonstrated that even
when behavioural scores fall within the normal range on standardised tasks, this does
not necessarily entail normal underlying cognitive processes (Grice, Spratling,
Karmiloff-Smith, Halit, Csibra, de Haan, & Johnson 2001; Karmiloff-Smith, 1997,
1998; see also the process/achievement distinction introduced by Werner, 1937).
Nonetheless, in the case of the previously normal adult patient and to the extent that
genuine cases of pure dissociations of behaviour indeed exist, then those data tend to
point to relative modularity of mind.
III.
Selective Deficits in Patients with Genetic Disorders
What about selective deficits in adults with genetic disorders? Do they suggest that
the mind starts off with independent modules that can be selectively spared or
impaired? First, it must be recalled that individuals with genetic disorders have not
developed normally to adulthood and then suddenly suffered a brain insult; their
brains have developed atypically from the outset. Yet, based on arguments from the
adult neuropsychological model, there exist numerous claims in the literature for
selective deficits in disorders of a genetic origin, against a background of intact or
preserved function, often allied to claims that such selective deficits are evidence that
cognitive modules are innate (Baron-Cohen, 1998; Leslie, 1992; Pinker, 1999; Smith,
1999; Smith & Tsimpli, 1995; Temple, 1997).
It should be noted that the terminology used here is misleading. While researchers
frequently employ terms like “intact” and “preserved” with respect to cognitive
functions in genetic disorders, this is not what they actually mean. The terms intact
Karmiloff-Smith & Thomas, page 6
and preserved imply that a pre-existing system has not been damaged. This may well
be appropriate for adults who have suffered damage to a previously normal cognitive
system, but there is no equivalent pre-existing state for the developmental disorder.
When researchers deploy terms like “intact”, “preserved”, or “normal” for a genetic
disorder, they are using shorthand, and they usually mean two things. First, they mean
that behavioural scores on certain standardised tests fall within the normal range. But,
second, they are proposing that the underlying cognitive processes have developed
normally. The unfortunate prevalence of this terminological shorthand has tended to
obscure the fact that claims about intact and impaired functions in genetic disorders
actually constitute implicit developmental theories.
More importantly, our view is that in developmental disorders, the existence of pure
deficits alongside so-called “intact modules” can be challenged not just empirically,
but theoretically as well. Even if selective, modular-like deficits were to exist in the
older child and adult, we argue that modules are the result of a process of
development, not its starting point (Karmiloff-Smith, 1992). The effects of a genetic
mutation during embryogenesis and postnatal brain growth are likely to be
widespread across the developing system. Some domains will be more affected than
others due to the different features of their particular problem space (Karmiloff-Smith,
1998; Karmiloff-Smith, Scerif & Thomas, 2002). It is crucial to take into account how
development itself might alter final outcomes. For instance, a tiny impairment in
infancy could impact differentially over developmental time, such that many domains
are affected but some display only very subtle deficits, whereas others are much more
obviously impaired. Some systems may develop atypically but still be able to generate
behavioural scores that fall within the normal range on coarse standardized
Karmiloff-Smith & Thomas, page 7
psychological tests. Since cognitive systems tend to interact during development
(either directly or indirectly via the environment), a cognitive system that is
developing atypically may begin to subtly perturb the developmental trajectories of
other systems with which it interacts, and particularly those that attempt to
compensate for emerging behavioural impairments. The complex dynamics of
interaction between neurocomputational constraints, the environment and ontogeny
cannot be ignored.
IV.
Constraints on Plasticity
Despite the above caveats about the interpretation of uneven cognitive profiles, it is
important to recognise that developmental disorders can actually be very informative
about constraints on plasticity. Plasticity is often invoked only in terms of the brain’s
response to hemispherectomies after epilepsy, to early or late damage and the like, as
if plasticity were a rare developmental process. As Wexler (1996) has put it:
"It is uncontroversial that the development [of Universal
Grammar] is essentially guided by a biological, genetically
determined program....Experience-dependent variation in
biological structures or processes...is an exception...and is called
'plasticity' “
Yet in our view, the notion of genetic determinism is misguided; plasticity is the rule,
not the exception (see, also, Cicchetti & Tucker, 1994, for a similar argument). Indeed,
plasticity is central to all development, normal or atypical, and explains how structure
changes as a function of experience. But plasticity is not, of course, unconstrained. The
initial properties of a learning system shape how change occurs following experience.
The adult brain is in fact the sculpted result of a complex interaction between the
Karmiloff-Smith & Thomas, page 8
individual’s processing of a wide variety of structured inputs and the
neurocomputational constraints on the developing brain. A crucial goal, therefore, is to
identify those constraints. But in identifying relatively consistent brain structures in the
normal adult, as in the normal child, it is difficult to distinguish two possible scenarios:
either the consistent structures emerge from consistent regularities in the environments
to which children are exposed – what some have called “species-typical environments”
(Johnson & Morton, 1991) – or they arise from consistent neurocomputational
constraints that shape the course of development despite variability in the environment.
It is here that experiments of nature may further our endeavours: some children start life
with brains that have atypical neurocomputational constraints but they live in relatively
normal environments, whereas other children may start out with potentially normal
brains but from the outset grow up in very atypical environments (in utero due to drugs
or alcohol, or in the outside world due to impoverished or violent conditions). This
distinction has also been discussed by Cicchetti (2002). The former case is the focus of
the present paper (for the latter – radically atypical environments – see review by
Thomas, in press). We submit that genetic developmental disorders can indeed inform
theories of normal development provided they are viewed as altered constraints on
neural plasticity in a developing organism and not as illustrations of “intact” versus
“damaged” static modules.
V.
Atypical Plasticity
As mentioned above, sometimes a specific deficit arises after a focal lesion to the
normal adult brain. However, as we shall shortly see, when a normal child suffers the
same focal lesions as an adult but in early development, selective deficits rarely
ensue. This is because normal processes of plasticity kick in and usually lead to
Karmiloff-Smith & Thomas, page 9
recovery from early insult, with no serious, lasting impairments. What about the case
of genetic developmental disorders without focal brain damage? Why doesn’t
plasticity simply compensate for the genetic mutation and allow the child to recover
function in similar ways to normal children with early focal brain damage? We
believe this is because of the improbability that the genetic mutation merely affects a
single, specific domain; rather, it is likely to affect plasticity itself, because genetic
mutations can give rise to atypical neurocomputational constraints from the outset
(Karmiloff-Smith, Scerif, & Thomas, 2002).
In this article, we use one genetic disorder, Williams syndrome (WS), to illustrate
how the atypical developmental pathways followed by this disorder may provide
clues to the normal neurocomputational constraints on development. We focus here
particularly on the domain of language acquisition, because it is WS language that
several authors have embraced as the perfect illustration of an intact grammar module
and the example of a dissociation between language and the rest of cognition (Pinker,
1991, 1994, 1999; Smith, 1999). Where appropriate, we contrast WS with normal
children who have experienced early brain damage and those suffering from other
developmental disorders of genetic origin. However, prior to doing so, we provide an
account of the WS genotype and phenotype in order to situate our subsequent
arguments.
VI. A Specific Example: Williams Syndrome
Williams syndrome (WS) is a rare neurodevelopmental disorder occurring in
approximately 1 in 20,000 live births (Morris, Demsey, Leonard, Dilts, Blackburn,
1988). It gives rise to specific physical, behavioural and cognitive abnormalities,
Karmiloff-Smith & Thomas, page 10
together with structural, chemical and functional anomalies in the developing brain
(Bellugi, Wang & Jernigan, 1994; Grice et al., 2001; Mervis, Morris, Bertrand &
Robinson, 1999; Rae, Karmiloff-Smith, Lee, Dixon, Grant, Blamire, Thompson,
Styles & Radda, 1998). The syndrome was initially reported by cardiologists (Beuren,
Apitz & Harmjanz, 1962; Williams, Barratt-Boyes & Lowe, 1961), who discovered
an association between the existence of supravalvular aortic stenosis (SVAS) and a
characteristic facial dysmorphology, together with growth retardation and learning
difficulties. Known as Williams syndrome in the USA and Williams-Beuren
syndrome in Continental Europe, it was initially called Idiopathic Infantile
Hypercalcaemia in the UK because some infants with a similar clinical description
also presented with infantile hypercalcaemia (Black & Bonham-Carter, 1963).
However, this turned out not to be a defining characteristic in subsequent cases, and
so the disorder is now also referred to as Williams syndrome in the UK.
A. The WS genotype
Most cases of Williams syndrome are sporadic, although a few instances of
concordant monozygotic twins and of parent to child transmission have been reported
in the literature (Morris, Thomas & Greenburg, 1993; Pankau, Gosch, Simeoni, &
Wessel, 1993). Although for some 20 years WS was diagnosed only on the basis of
clinical criteria, the early 1990s hailed the discovery of the first genetic markers of the
syndrome (Curran, Atkinson, Ewart, Morris, Leppert, & Keating, 1993; Ewart,
Morris, Ensing, Loker, Moore, Leppert, & Keating, 1993a). Curran and collaborators
pointed to an association between a disruption due to the translocation of one copy of
the elastin gene at chromosome 7q11.23 and the existence of supravalvular aortic
stenosis, a common feature of WS. Ewart and colleagues then confirmed that
Karmiloff-Smith & Thomas, page 11
hemizygosity at the elastin locus also occurred in patients with WS (Ewart, Morris,
Atkinson, Weishan, Sternes, Spallone, Stock, Leppert, & Keating, 1993b). The elastin
gene is expressed during the third trimester in utero and during early postnatal life. Its
deletion leads to the production of structurally abnormal tropelastin and causes
problems with elasticity and connective tissue in numerous parts of the organism,
particularly the skin and arteries.
The microdeletion measures some 1.5Mb and is of fairly uniform size in the majority
of patients with WS, with the elastin gene being midway between the two breakpoints
(Perez-Juralo, Peoples, Kaplan, Hamel & Franke, 1996). Parental origin seems to play
no role, with deletions on the maternally and paternally inherited chromosomes
occurring with equal frequency. The mutational mechanism seems to lay in unequal
meiotic recombination between chromosome 7 homologues, although
intrachromosomal rearrangements also arise. More recently, it has been shown that
large repeats containing genes and pseudogenes flank the two deletion breakpoints,
thereby lending themselves to chance misalignment (Peoples, Franke, Wang, PerezJurado, Paperna, Cisco, & Franke, 2000; Robinson, Waslynka, Bernasconi, Wang,
Clark, Kotzot, & Schinzel, 1996; Urban, Helms, Fekete, Csiszar, Bonnet, Munnich,
Donis-Keller & Boyd, 1996).
Once the locus of the WS deletion was documented, several others genes started to be
identified in the critical region. Much excitement was generated by the discovery of
the deletion of the LIM Kinase-1 gene because it is expressed in the brain. LIMK1 is
situated telomeric to ELN and was found to be deleted in all patients with a typical
WS deletion (Frangiskakis, Ewart, Morris, Mervis, Bertrand, Robinson, Klein,
Karmiloff-Smith & Thomas, page 12
Ensing, Everett, Green, Proschel, Cutowski, Noble, Atkinson, Odelberg & Keating,
1996; Tassabehji, Metcalfe, Fergusson, Carette, Dore, Donnai, Read, Proschel,
Gutowski, Mao & Sheer, 1996). LIMK1 encodes a protein tyrosine kinase that
inactivates cofilin, a protein required for the turnover of actin filaments (Arber,
Barbayannis, Hanser, Schneider, Stanyon, Bernard & Caroni, 1998). Mouse models
had already shown that during embryogenesis LIMK1 is expressed in the central
nervous system, including the inner nuclear layer of the retina, the cortex, the spinal
cord, the cranial nerves and dorsal root ganglia (Proschel, Blouin, Gutowski, Ludwig,
& Noble, 1995). Thus, mutations in the expression of LIMK1 in Williams syndrome
are likely to affect axonal guidance during the crucial building of the central nervous
system (for a review of the genes in the WS critical region, see Franke, 1999).
B. The WS physical phenotype
Mean birth weight in WS is low, with postnatal growth retardation frequently
reported. Early puberty also often contributes to the low final adult height (Cherniske,
Sadler, Schwartz, Carpenter & Pober, 1999). The facial dysmorphology is particularly
characteristic. Patients have a flat nasal bridge and anteverted nares, a wide mouth
with fleshy lips, periorbital fullness, flat malar region, small mandible and prominent
cheeks. Failure to thrive, often due to difficulties in sucking, is reported during early
infancy. A prematurely aged appearance is often apparent in late adolescence and
early adulthood. As mentioned, supravalvular aortic stenosis and peripheral
pulmonary artery stenosis are very common in the syndrome (Hallidie-Smith &
Karas, 1988), with elevated blood pressure also frequently noted in adolescents and
adults (Broder, Reinhardt, Ahern, Lifton, Tamborlane & Pober, 1999). Some 50% of
children with WS have strabismus, commonly with refractive errors (Atkinson,
Karmiloff-Smith & Thomas, page 13
Anker, Braddick, Nokes, Mason & Braddick, 2001; Winter, Pankau, Amm, Gosch &
Wessel, 1996). Hyeracusis – an unusual hypersensitivity to certain sounds – is present
in as many as 95% of patients, whereas audiometry is usually normal. Little is yet
known about central auditory processing in Williams syndrome although new
research is underway (Cohen, Ansari, Rosen & Karmiloff-Smith, 2002).
C. The WS behavioural and cognitive phenotypes
In experimental psychology and cognitive neuroscience, the terms “behavioural” and
“cognitive” have different meanings from the way in which they are employed in
clinical medicine. For the clinician, “behavioural phenotype” refers to emotional and
personality traits, attention deficits, as well as IQ scores. “Cognitive” is simply
subsumed under the term “behavioural” and refers to the standardised measures of
intelligence. In experimental psychology, by contrast, “behavioural” refers to
measures of overt behaviour, e.g., scores relating to whether or not the subject
succeeds or fails at a task. The “cognitive” level attempts to account for the mental
processes underlying the overt behaviour, i.e., how the individual processes the inputs
involved in the task. This is not unsimilar to an early but hitherto rather neglected
distinction made by Werner (1937) between achievement and process. It must be
always recalled, therefore, that the same overt behaviour may be achieved by different
underlying cognitive processes when comparing various disorders to the normal case
(Karmiloff-Smith, 1998). Thus, claims that certain aspects of a profile are
“unimpaired”, “intact”, “preserved” and the like, need to be taken with caution until
the cognitive level has been fully explored. To understand a syndrome in any depth, it
is vital to distinguish between the behavioural phenotype and the cognitive
phenotype.
Karmiloff-Smith & Thomas, page 14
The behavioural phenotype in WS includes what has been termed ‘hypersociability’
(Jones, Bellugi, Lai, Chiles, Reilly, Lincoln & Adolphs, 2000), with a tendency to be
overly friendly with strangers and to lack social judgement skills (Einfeld, Tonge &
Florio, 1997; Gosch & Pankau, 1997). People with WS frequently display high
anxiety about new situations. They are, however, particularly empathetic towards
others’ emotions, but far less skilled at understanding human intentionality (TagerFlusberg, Boshart & Baron-Cohen, 1998). Most studies estimate a mean full IQ of
between 51-70 (Bellugi, et al., 1994; Mervis et al., 1999; Udwin & Yule, 1991). To be
noted, however, is the fact that full IQ scores camouflage marked unevenness in the
WS profile, with verbal IQ outstripping performance IQ in the majority of cases. The
pioneering work of Bellugi and her collaborators had originally suggested some clearcut dissociations in the cognitive architecture of WS. Language and face processing
appeared to be “preserved” alongside both general retardation and particularly serious
problems with visuo-spatial cognition (Bellugi, et al., 1994). However, as mentioned
above, the notion that behavioural abilities in a developmental disorder directly index
underlying cognitive processes is one that we challenge (Karmiloff-Smith, 1998;
Thomas & Karmiloff-Smith, in press, a). Indeed, in-depth analyses of the language
and face processing of WS adults – two areas frequently reported to be “intact”–
strongly suggest that the behavioural proficiencies of these individuals are supported
by different cognitive processes compared with normal controls.
Karmiloff-Smith & Thomas, page 15
VI.
Neurocomputational constraints on development in Williams
syndrome
Having set the stage for understanding the characteristics of our particular experiment
of nature – Williams syndrome – we can now turn to the issues laid out in our
introductory sections as to what developmental disorders can tell us about the
neurocomputational constraints that shape development, using language acquisition as
our main illustration. The rest of this article will proceed as follows. We begin with a
consideration of the (static) endstate of language development in WS, including
claims for selective deficits and dissociations. To emphasise the role that alteredconstraints-on-plasticity must play in WS, we identify the type of acquired damage
that causes similar patterns of behavioural deficits in normal adults, and then examine
what happens to normal children who experience such damage early in development.
As we have seen, in our target disorder the cause is not acquired damage but a genetic
mutation acting from embryogenesis onwards. Consideration of a realistic causal link
between (potentially mutated) genes and (potentially altered) behaviour leads us to
focus on the precursors of language development in WS. It is here that we have the
best opportunity to reveal the intrinsic neurocomputational constraints that differ in
the infant with WS, before they are obscured behind layers of development driven by
dynamic interaction with the environment. A review of the available evidence leads
us to a hypothesis concerning the constraints that may be atypical in WS language
(and by implication, that must act appropriately in normal development). Armed with
these, we offer a developmental theory of the acquisition of language in this disorder
and how this might enlighten theories of typical development. Finally, we focus on
how one might explore more precisely the influence of constraints on development
Karmiloff-Smith & Thomas, page 16
within a neurocomputational framework via the application of connectionist
modelling techniques.
A. Language in late childhood/adulthood in Williams syndrome
In the initial characterisations of Williams syndrome, it was thought that the seeming
dissociation between language and other general reasoning abilities might represent
an informative ‘experiment of nature’. Thus, Rossen, Klima, Bellugi, Bihrle & Jones
(1996) commented that the syndrome ‘presents a remarkable juxtaposition of
impaired and intact mental capacities…[..]...linguistic functioning is preserved in
Williams syndrome while problem solving ability and visuospatial cognition are
impaired’. In Pinker’s initial references to the disorder (1991, 1994), he viewed WS as
an example of the potential developmental independence of language and cognition,
consistent with his theory that language (and particularly syntax) is an innate, selfcontained module. Indeed, he contrasted WS with Specific Language Impairment
(SLI), a disorder that presents as a case of impaired language development with
apparently normal non-verbal cognition and intelligence. Given that SLI has a
heritable component, Pinker (1999) argued that WS and SLI together represent a
‘genetic double dissociation’ pointing to the developmental independence of language
from cognition.
However, subsequent careful research has led researchers to more refined claims,
given that a range of studies has revealed anomalies at all levels of language in WS,
including phonology, lexical-semantics, grammar, and pragmatics. Such studies
include the examination of WS populations in several countries, such as the USA,
Britain, Italy, France, Spain, and Germany (see Thomas & Karmiloff-Smith, in press,
Karmiloff-Smith & Thomas, page 17
a, for a review of this work). It does appear true that in some cases, individuals with
WS are competent at understanding and producing a wide range of sophisticated
grammatical constructions (see, e.g., Clahsen & Almazan, 1998). (It should of course
be kept in mind that, as in many developmental disorders, there is great individual
variability in WS. Some lower functioning individuals show much poorer language
despite having the same genetic deletion.) On the other hand, in almost all cases of
WS, language performance falls below that found in chronological age-matched
controls. Indeed, most studies of language in WS now compare performance against
mental age controls. Paradoxically, those claiming that language and cognition are
independent also opt for MA-matched controls; yet this very choice implicitly accepts
a relationship between language and cognition! And it is worth noting that full IQ
measures make the language results look more impressive than they really are. To
state that a 25-year-old adult with WS has a full IQ of 53 but understands complex
embedded clauses seems impressive. But assessed from a different angle, that of
Mental Age, it is much less amazing. The same 25-year-old individual with WS may
have a Verbal Mental Age of 8, but then there is nothing surprising about the fact that
typically developing 8-year-olds have mastered many complex, structural aspects of
language (see discussion in Karmiloff-Smith, 1998). It is often the pragmatic
limitations of WS language that draw listeners’ attention to the fact that this
individual’s conversation is not at the level we expect for a typical 25-year-old, or an
8-year-old, for that matter. For example, in WS, speech content is often odd or out of
place in a particular social context (Volterra, Capirci & Caselli, 2001), speech can
contain high levels of clichés and stereotyped phrases (Howlin, Davies & Udwin,
1998a), and the comprehending of non-literal language, can be very deficient
(Howlin, Davies & Udwin, 1998b).
Karmiloff-Smith & Thomas, page 18
Despite a retreat from some of the stronger claims, two aspects of language in WS
remain notable. First, the language of these individuals is often more advanced than
that found in other genetic developmental disorders with comparable IQ, such as
Down’s syndrome (Bellugi, Wang & Jernigan, 1994; Mervis et al., 1999). Second,
language performance in WS is markedly out of step with their visuo-spatial
cognition, which is particularly poor and shows a characteristic ‘featural’ style of
processing (Deruelle, Mancini, Livet, Cassé-Perrot, & de Schonen, 1999; Donnai &
Karmiloff-Smith, 2000). Indeed, one of the defining criteria of the cognitive
phenotype of WS is the disparity between performance on standardised tests of
vocabulary and on standardised tests of visuo-spatial construction (such as a copying
a pattern with a set of coloured blocks, or copying a picture) (Mervis, Robinson,
Bertrand, Morris, Klein-Tasman, & Armstrong, 2000).
However, claims persist that selective deficits can be found within the language
systems of individuals with WS, and that such deficits can inform theorists about the
structure of the normal language system. For instance, Temple and colleagues argue
that ‘the linguistic performance of [individuals with] WS can be explained in terms of
selective deficits to an otherwise normal modular system’ (Temple & Clahsen, in
press, italics added), and as a pattern of ‘some preservation and some disability’
(Temple, Almazan, & Sherwood, 2002). The latter claims for deficits have focused on
the representation of and access to word-specific knowledge. For instance, the
language of individuals with WS has frequently been characterised as containing
unusual or low frequency items (e.g., Bellugi et al., 1994; Pinker, 1991, 1994, 1999;
Rossen et al., 1996; Temple et al., 2002), and older children with WS have been
Karmiloff-Smith & Thomas, page 19
argued to demonstrate naming deficits and difficulties representing fine semantic
distinctions (Temple et al., 2002). In the domain of inflectional morphology, it has
been argued that word inflections involving word-specific knowledge (such as
irregular past tenses and plurals) are selectively impaired in WS, while those
following grammatical rules are in line with mental age (Clahsen & Almazan, 1998,
2001; Pinker, 1991, 1994), but, note, not chronological age. Overall the claim has
been that grammar is preserved in WS (i.e., develops normally), while some aspects
of word-knowledge are impaired (i.e., develop atypically). To reiterate, the
comparison is with mental age, not chronological age, so such claims simply dismiss
the considerable delay as if delay in one system were completely independent of the
rest of the developing system. WS has none the less been used to support the
position that there is a dissociation between language and cognition and that the
distinction between grammar and the lexicon is innate (e.g., Pinker, 1999).
There are a number of shortcomings with these arguments, not least the fact that many
of the empirical findings on which they are based have employed very small numbers
of participants with WS and that they have often failed to replicate when larger
samples have been used (e.g., standardised tests of word fluency reveal that WS
vocabulary is no more atypical than mental age controls: Jarrold, Hartley, Phillips &
Baddeley, 2000; Scott, Mervis, Bertrand, Klein, Armstrong, & Ford, 1995; there is no
significant selective deficit for irregular inflections over regular inflections: Thomas,
Grant, Gsödl, Laing, Barham, Lakusta, Tyler, Grice, Paterson & Karmiloff-Smith,
2001, and Zukowski, 2001; naming skills are in line with verbal mental age: Thomas,
Dockrell, Messer, Parmigiani, Ansari & Karmiloff-Smith, 2002). However, the
selective-deficit claims are illustrative of the view that it is sufficient to explain a
Karmiloff-Smith & Thomas, page 20
developmentally disordered system with reference to a static model of the normal
adult language system, and to orient an empirical program merely towards identifying
those components that are deficient – thereby negating altogether the contribution of
development.
We argue that this appeal to static models has inhibited progress in understanding WS
in particular and developmental disorders in general, but we do not wish to overstate
this case. Even in the more static accounts, with their postulation of preserved and
impaired components, there are occasional hints among researchers that the final
explanation of anomalies in the WS system must fall within a developmental
framework. Thus, Pinker comments that ‘presumably LIM-kinase1 [deleted from one
copy of chromosome 7 in WS] plays an important role in the development of the
neural networks used in spatial reasoning, possibly in the parietal lobes. The other
missing genes, perhaps, are necessary for the development of other parts and
processes of the brain, though not for language’ (1999, p.260-1, italics added). When
Rossen et al. (1996) appealed to a static model of the WS lexicon to explain the
presence of unusual words in WS vocabulary, they sought to characterise the problem
as a specific anomaly of activation dynamics against a background of normal lexical
structure: ‘… While individuals with WS have well organised semantic categories,
and have good access to word knowledge, an anomaly does exist in some tasks
dependent on consideration of words in a complex activation environment’. Yet in the
same article, when these researchers began to consider the impact of development on
the disorder, they added that, so long as activation dynamics were involved in
knowledge acquisition, knowledge could not be normal while activation dynamics
were anomalous: ‘Contemporary neural models of learning postulate definable
Karmiloff-Smith & Thomas, page 21
quantitative relationships bridging the classical dichotomy of structure and process…
Within this model paradigm, if activity malfunction with the lexicon exists, then
abnormality of lexicon structure will likely follow’ (Rossen et al., 1996).
Nevertheless, the prevalence of theories of the “preserved/impaired” kind provides an
opportunity to stress our alternative claim that genetic developmental disorders are
best viewed as altered constraints on plasticity. To make this argument, we first
consider what type of brain damage might underlie the types of selective deficits
proposed, were they to appear in a normal adult. We then turn to the literature on the
developmental outcomes of these types of brain damage when they occur in typically
developing young children, in whom we assume that normal constraints on plasticity
hold.
The clearest parallels between behavioural deficits in WS and those found in normal
adults with acquired brain damage have been made in the domain of word inflections
(e.g., past tense of verbs, plurals of nouns, comparatives of adjectives). Thus Pinker
(1991, 1994) has explicitly argued that the deficits in inflectional morphology found
in WS are similar to those occurring in fluent aphasia in normal adults, following
damage to left temporal areas. These patients also show difficulties in word retrieval.
Similar parallels are drawn with deficits displayed in neurodegenerative disorders,
such as Alzheimer’s disease and cases of semantic dementia. These disorders are also
characterised by acquired functional damage that is greater in the temporal areas and
adjacent parietal areas than in the frontal lobes. And for reference, as the earlier
quoation from Pinker implies, visuo-spatial constructive deficits are found in normal
Karmiloff-Smith & Thomas, page 22
adults following damage to parietal areas, and in particular, the featural style of
processing tends to be associated with right parietal damage.
Note, however, that individuals with WS do not suddenly suffer discrete brain damage
in adulthood, but have anomalies in brain structures from the start of brain
development. What happens if these types of focal brain damage – to left temporal
areas, or to right parietal areas – occur to normal individuals early in infancy?
Another set of nature’s experiments can provide the answer.
B. Evidence for atypical constraints on plasticity in Williams syndrome
Bates and Roe (2001) recently reviewed the evidence concerning the effects of early
unilateral brain damage on the language development of young children, including the
effect of lesion side (left or right hemisphere) and lesion site. There were three main
findings. First, almost all the brain injured children exhibited delays in first-word
production, regardless of lesion side or site. Bates and Roe interpret these data as
suggesting that it is hard to get language “off the ground” after significant damage to
either hemisphere. Second, delays in word production tended to be more severe in
children with left posterior damage, particularly in the temporal lobe. Bates and Roe
note the apparently greater importance of the left hemisphere, but interestingly the
fact that in the adult, it is left frontal not temporal posterior damage that is associated
with expressive language deficits. Third, and most relevant to our concerns, is the fact
that the evidence suggests that when damage occurs prior to around 5 to 7 years of
age, plastic reorganisation takes place such that when tested later, these children show
little if any language impairment. Importantly, there are no broad effects of side of
Karmiloff-Smith & Thomas, page 23
damage. In short, early damage to left (or right) hemisphere when there is normal
plasticity tends to lead to recovery, with no long-lasting gross deficits.
The implication is that the analogies drawn between behavioural deficits after focal
lesions in normal adults and behavioural deficits in individuals with WS do not point
to a common underlying cause. Where focal damage occurs in early childhood,
normal plasticity is sufficient to effect behavioural recovery as a result of ontogenetic
development. If we rule out focal damage + normal plasticity as an explanation for
WS, then several options remain. There could be widespread damage + normal
plasticity, focal damage + atypical plasticity, or widespread damage + atypical
plasticity. Of these options, increasing evidence suggests that the brains of individuals
with WS do not present with focal lesions (and this is also the case for almost all other
genetic developmental disorders), but with widespread differences in neuroanatomical
features (Bellugi, Wang & Jernigan, 1994), in neuronal density (Galaburda, Wang,
Bellugi & Rossen, 1994), in biochemistry (Rae et al., 1998), and in
electrophysiological activity (Grice et al., 2001; Neville, Mills & Bellugi, 1994).
The possibility that WS is caused by widespread but subtle damage + normal
plasticity merits consideration. Rourke (e.g., 1987, 1989) proposed the notion of nonverbal learning disability to account for a consistent pattern of behavioural deficits
observed in children with a history of early, generalised cerebral deficits. Anderson,
Northam, Hendy and Wrennall (2001) summarise the main features as follows:
bilateral tactile-perceptual deficits, more marked on the left side of the body, impaired
visual recognition and discrimination, impaired visuospatial organisation, bilateral
psychomotor coordination problems, again more marked on the left side of the body,
Karmiloff-Smith & Thomas, page 24
and difficulties managing novel information. Among these weaknesses are also areas
of relative strength, which include simple motor skills, auditory perception, rote
learning, selective and sustained attention for auditory-verbal information, basic
expressive and receptive language, and word reading and spelling. Non-verbal
learning disorder bears a number of similarities to the features of WS, in particular the
relative strength of language compared to the deficits in visuospatial skills. Indeed,
within language itself, Rourke and Tsatsanis (1996) point to a dissociation between
better performance on structural aspects of language than on pragmatics, another
similarity with WS. However, differential perceptual and motor problems with the left
side of the body have never hitherto been reported in WS.
Rourke’s notion of non-verbal learning disability (NVLD) is interesting here in that it
is allied to an underlying neurological explanation called the ‘white matter
hypothesis’ (Rourke, 1987; see Anderson et al., 2001, for discussion). The idea is that
the global connectivity of the brain, the white matter, has a prolonged period of
development, in which it is particularly vulnerable to disruption after widespread
damage in children who have experienced conditions such as traumatic brain injury,
cranial irradiation to treat tumours, or hydrocephalus. (Note that under some
circumstances, these conditions may be associated with a normal genotype, such that
one might expect normal plasticity if the mechanisms that support it have survived.)
Rourke then distinguishes between the roles of white matter in the left and right
hemispheres. In the right hemisphere, he proposes that white matter is important for
both development and maintenance of skills (including global integration of
information), while in the left hemisphere it is important for the development of skills
but not necessarily their maintenance. In this view, the right hemisphere is responsible
Karmiloff-Smith & Thomas, page 25
for more integrative functions and the left hemisphere for more segregated functions.
Widespread damage to the white matter then results in greater impairments to the
functions of the right hemisphere (such as visuo-spatial integration and pragmatic
aspects of language), than to the more self-contained functions of the left hemisphere,
(such as processing of visual detail, auditory processing, basic expressive and
productive language).
It is an appealing notion that that widespread damage can itself result in differential
(seemingly selective) effects on the functions of left and right hemisphere, which
normal plasticity is unable to overcome during atypical development (due to the
severity of the damage), for this would explain the subsequent uneven profile of
cognitive abilities. However, it is far from clear than NVLD is a homogeneous
syndrome – indeed, Rourke (1987) prefers to conceptualise it as a ‘final common
pathway’ of deficits that may be caused by a range of underlying pathologies. A range
of disorders, including both those of genetic origin and of acquired early damage, are
subsumed under this heading, each of which shares different levels of similarity to the
prototypical description of the disorder. As such, NVLD may be a label attached to
the common behavioural deficits shared by very sub-optimal cognitive systems,
systems which differ as to their underlying causes and which, at a detailed level, differ
in their cognitive symptoms. Gross behavioural deficits in the more challenging,
integrative aspects of cognition may then be reached either by a system with normal
plasticity but widespread damage (if indeed the mechanisms of normal plasticity
remain after such damage), or by a system with both widespread anomalies and
atypical plasticity. It is the latter possibility that we would argue obtains in the case of
genetic developmental disorders such as WS, for reasons that we address in the next
Karmiloff-Smith & Thomas, page 26
section. Nevertheless, despite the common tendency to seek dissociations, the
behavioural similarities across genetic syndromes can also be very informative
(Bishop, 1997; Karmiloff-Smith, 1997, 1998). Given the complex and indirect
relationship between genes and brain development, it remains possible that different
genes have similar distal effects on brain development and lead to similar cognitive
deficits. Indeed, such a possibility has been explicitly raised regarding similarities
identified between WS and velocardiofacial / DiGeorge syndrome (Beardon, Wang, &
Simon, 2002), although this initial comparison has yet to be established in depth.
C. A causal model for exploring the relationship between genes and
behaviour
We have argued that the effects of a genetic mutation that emerge during
embryogenesis and postnatal brain growth are likely to be widespread across the
developing system, albeit relatively subtle, but that a consequence of the mutation can
be the emergence of an uneven cognitive profile in the endstate. In trying to link the
genetic mutation to the final behavioural profile, it is important to have a realistic
causal model in mind that could, however indirectly, connect the two.
Despite the fact that in recent times (and often, admittedly, in the popular press),
genes have been specifically linked to behaviours (such as the “gene for language”,
the “gene for crime”, the “gene for homosexuality”, and so forth), the link between
genes and adult behaviour is of course incredibly indirect and involves many-to-many
rather than one-to-one mappings between genes and behavioural phenotypes.
Moreover, genetic mutations are limited in the ways in which they can disrupt brain
development. Pennington (2001) suggests three broad classes of effects, those on (1)
Karmiloff-Smith & Thomas, page 27
brain size, in terms of altering the number of neurons or synapses, (2) neuronal
migration, sometimes in a regionally specific fashion, and (3) neurotransmission,
either by changing levels of neurotransmitter or changing the binding properties of
receptor proteins. To this we would add the fact that genetic mutations can also affect
the timing of gene expression which will have cascading effects on the developing
organism, since timing is a crucial aspect of the emergent organisation of the
functional structure (Elman, Bates, Johnson, Karmiloff-Smith, Parisi & Plunkett,
1996; Huttenlocher, 2002; Karmiloff-Smith, 1992, 1998). Importantly, these effects
impact on the neurocomputational properties of the brain, including the way in which
it modifies its structure as a consequence of internally- or externally-generated
activity. Subsequent behavioural deficits are the outcome of a long trajectory of
development, whereby cognitive structures emerge via an interaction between internal
neurocomputational constraints and the environment.
In placing development itself as a vital factor in the explanation of developmental
disorders (Karmiloff-Smith, 1998), we have characterised the developmental
emergence of cognitive structures as a type of recursive equation (Thomas &
Karmiloff-Smith, in press, a). The equation is as follows: More complex cognitive
structure = Less complex structure X process of development. This means that the
process of development corresponds to the interaction of an internal or external
environment with the existing neurocomputational constraints1.
In atypical development, the neurocomputational constraints can differ. Under these
conditions, each cycle of recursion that operates on a system may produce functional
properties that fall inside or outside the normal range, and for any subsequent steps of
Karmiloff-Smith & Thomas, page 28
development that rely on this system, a new set of constraints that are more or less
atypical for the next pass through the equation. The result is that across the whole
developing cognitive system, there emerges a complex, graded pattern of areas that
function either within or outside the range of variation that one might expect in the
normal population. (In cognitive terms, these graded variations are with respect to the
functional properties of sub-systems rather than their purely anatomical or
physiological properties). Whether a given subsystem falls inside or outside the
normal functional range as a consequence of the operation of the equation depends on
the extent to which the development of that subsystem relies upon the particular
neurocomputational constraints that were the atypical outcome of the previous
operation of the equation2. Ultimately, this cascade of equations emanates from the
gene products that were altered in embryogenesis by the original genetic mutation.
This is a complex formulation, but we believe it is a more realistic causal pathway
than that offered by static theories calling on preserved/impaired modules, because it
includes development at every stage. Importantly, as a causal model, it seeks to
remain at a single (in this case, cognitive) level of description. Causal relations
cannot, in our view, operate across levels of description. That is, neuronal activations
do not cause mental representations, they are mental representations. Genes cause
molecular events. Thus genes are not only separated from behaviour by a long and
complex developmental trajectory, but also by existing at a different, lower level of
description.
Having taken this point on board, we are presented with a problem. If, to identify the
causal origins of cognitive deficits, we must remain at the cognitive level of
Karmiloff-Smith & Thomas, page 29
description, then we are faced with the difficulty that the foetus and newborn infant
do not have cognition in the way that children or adults do. To construct the causal
account, we must therefore describe a proto-cognitive system in the newborn or infant
that produces behavioural precursors to later complex behaviours that we see in
children and adults (Karmiloff-Smith, 1998). With development, mental
representations begin at a simple, concrete level and achieve an increase in
abstractness and complexity following operation of the recursive equation we have
proposed. Now, some might describe more complex/abstract cognitive processes as
“higher-level” functions, and in this sense, the equation recursively climbs from lower
to higher levels of representation across development. But crucially, the distinction
between low-level and high-level representations is one of complexity and
abstractness, while the causal relations remain at the same cognitive level of
description. (Of course, different disciplines will seek to characterise this recursive
causal equation at other levels of description, such as at the level of neural structure
when characterising the emergence of atypical brain structures).
We have taken a short diversion from WS to make this conceptual clarification, but
the outcome is a clear message of methodology. If we are to tease apart the cascade of
developmental equations that has led to cognitive deficits in the adult state, we must
pursue those deficits back down the cascade, to their origins in infancy (Brown,
Johnson, Paterson, Gilmore, Longhi, & Karmiloff-Smith, 2003; Karmiloff-Smith,
1998; Paterson, Brown, Gsödl, Johnson & Karmiloff-Smith, 1999). In studying
cognitive precursors of complex behaviours, we will find the clearest evidence of the
altered neurocomputational constraints that are the first cognitive consequence of a
genetic mutation.
Karmiloff-Smith & Thomas, page 30
D. Atypical precursors to language in Williams syndrome
The most salient aspect of the onset of language in WS is that it is delayed. Although
this delay is variable, one study of 54 children with WS found an average delay of 2
years, similar to that found for children with Down syndrome (DS) (Singer Harris,
Bellugi, Bates, Jones & Rossen, 1997; see also Paterson et al., 1999). Though
delayed, some aspects of early development reveal normal behavioural patterns. For
example, the onset of hand banging predicts the onset of canonical babbling in infants
with WS in the same way as it does in typically developing infants (Masataka, 2001;
Mervis & Bertrand, 1997). And, once language development commences, the
relationship between vocabulary size and grammatical complexity, although seriously
delayed, is again within the normal range, a pattern that contrasts with DS where
grammatical complexity is reduced for a given vocabulary level (Singer Harris et al.,
1997).
Despite the fact that phonological memory appears as a relative strength in WS in
childhood and adulthood (Mervis et al., 1999), a study of the ability of infants and
toddlers with WS to segment fluent speech stream into words revealed serious delays
(Nazzi, Paterson, & Karmiloff-Smith, 2003). In part, then, language delays may be
due to problems with the early development of speech perception and phonological
representations.
However, some precursors appear not just delayed, but atypical. For example, Laing
and colleagues examined socio-interactive precursors to language development in
toddlers with WS compared with MA controls (Laing, Butterworth, Ansari, Gsödl,
Karmiloff-Smith & Thomas, page 31
Longhi, Panagiotaki, Paterson & Karmiloff-Smith, 2002). Although toddlers with WS
were proficient at dyadic interactions with a caregiver (and indeed sometimes
exceeded the scores of MA controls due to persistent fixation on the caregiver’s face;
see also Bertrand, Mervis, Rice & Adamson, 1993; Jones et al., 2000), there was a
marked deficiency in triadic interactions incorporating an object. Specifically,
toddlers with WS had difficulty switching attention from the caregiver to an object
that was being referred to in communication (via pointing, looking, and naming). One
might imagine that this deficiency would disadvantage the toddlers with WS in
learning the names of objects, since shared attention to newly named objects appears
one of the main routes into vocabulary acquisition. And indeed, there is accumulating
evidence that precursors to vocabulary development in WS are atypical.
Typically developing infants use the presence of linguistic or gestural information
accompanying the introduction of novel objects to influence their subsequent
categorisation of those objects, sometimes over and above the perceptual similarities
among the objects. However, Nazzi and Karmiloff-Smith (2002) found that 2-6 year
old children with WS were significantly less able than typical controls to use verbal
cues to constrain categorisation. Masataka (2000) found a similar poverty in the
ability of 2-3 year olds with WS to use gestural information to constrain
categorisation.
In typically developing children, the ability to use pointing to refer to objects tends to
emerge before the use of verbal labels for the same purpose. Presumably, pointing
indexes the emergence of the cognitive ability to make reference, prior to the lexical
manifestation. Pointing to objects and eliciting pointing behaviour in adults also
Karmiloff-Smith & Thomas, page 32
facilitate the ability to find the correct referent for a given label. However, in WS,
Mervis and Bertrand (1997) found that the order was reversed, with the onset of
productive vocabulary preceding pointing. Laing et al. (2002) confirmed a deficit in
the pointing behaviour of infants with WS, despite relative proficiency at fine motor
skills. Vocabulary acquisition, therefore, appears to rely on a different set of cues and
constraints in WS. When Stevens and Karmiloff-Smith (1997) examined the
constraints that older children and young adults with WS were using to learn labels
for novel words, these, too, appeared atypical.
It is often the case that WS seems to present with a mixture of typical and atypical
patterns within the same domain. For example, Mervis and Bertrand (1997) found
that when playing with toys, non-verbal play patterns and object label comprehension
patterns in children with WS showed the normal priority of basic level categories over
subordinate categories, a pattern also found in DS. However, relations between
markers of semantic knowledge and productive vocabulary appeared yet again
atypical in the WS group. For instance, spontaneous exhaustive sorting of objects
(such as arranging toy animals and blocks into their separate categories) indexes the
development of semantic knowledge, and tends to precede a rapid rise in the rate of
vocabulary acquisition in typically developing children. Thus, by the time children
find it clear which categories objects fall into, it becomes increasingly easier for them
to attach consistent labels to different objects. However, for children with WS, Mervis
and Bertrand (1997) found no evidence that exhaustive sorting preceded the
vocabulary spurt; indeed, several children with WS exhibited the reverse pattern
(unlike children with DS, who displayed the normal pattern).
Karmiloff-Smith & Thomas, page 33
Finally, anecdotal parental reports have suggested that young children with WS
sometimes appear to say more than they actually comprehend (Singer Harris et al.,
1997). Consistent with this, there is some systematic evidence that compared to
normal children, the vocabulary of young children with WS exhibits a reduced
advantage for comprehension vocabulary over production vocabulary (Paterson,
2000), implying a relatively higher productive vocabulary for their level for
comprehension.
In summary, precursors to language development in WS paint a picture with two main
themes. First there is an overall delay, perhaps of a more generalised nature
incorporating delays in at least motor, phonological, and semantic development.
Second, when language development gets underway, a differential balance emerges
between the ability to encode and produce word forms on the one hand, and the
acquisition of the semantic underpinnings for those words on the other.
E. A developmental framework for language acquisition in Williams
syndrome
The psychological literature on the language of older children, adolescents, and adults
with WS is more substantial, but two types of hypothesis can be identified within it
(Thomas & Karmiloff-Smith, in press, a). The first hypothesis is a conservative one,
that the language of individuals with WS is broadly in line with their overall learning
disabilities. From this perspective, some anomalies arise as an indirect effect of their
other deficits, such as the visuo-spatial deficit which causes a difficulty with acquiring
prepositions that encode spatial relations (Jarrold, Phillips, Baddeley, Grant &
Karmiloff-Smith, 2001). Interestingly, the data also point to a general problem in WS
Karmiloff-Smith & Thomas, page 34
with relational terms like “darker than” that are non-spatial and yet involve the need
to generate spatialized internal mental representations to process them. However,
although this may explain local difficulties in some aspects of WS language, such a
conservative hypothesis fails of course to account for why language development
should be better in WS than in other disorders with comparable IQ, such as DS.
The second hypothesis is the more prevalent and, in line with the conclusions of the
preceding section, views language in WS as evolving according to an atypical balance
of constraints from phonology and semantics. As yet, there is no consensus on
whether the differential balance is caused simply by a relative strength in phonology,
simply by a relative weakness in semantics, by a difficulty in integrating these two
sources of information, or some combination of these possibilities. Nevertheless, there
is now sufficient evidence to begin to sketch out a developmental theory of language
acquisition in WS, a theory that stands in marked contrast to accounts that have dealt
purely in terms of selective deficits to a static model of the normal system. We have
constructed this preliminary account in the context of our recent work exploring the
causes of the unusual, socially-engaging vocabulary in the language of individuals
with WS (Thomas et al., 2002). This phenomenon is one of its most widely reported
characteristics, and a finding that led to the claims about isolated deficits in wordspecific knowledge that we discussed in a previous section.
In our preliminary account, language in WS is seen in a wider socio-communicative
context, and in a developmental framework that takes into consideration the way in which
compensation can occur in a system that has different initial processing biases but can be
adapted to meet the social needs of the individual. From this perspective, the unusual
Karmiloff-Smith & Thomas, page 35
characteristics of conversation in WS reflect a form of developmental compensation in
which language is used primarily to meet the (possibly heightened) social needs of the
individual (Jones et al., 2000), but acquisition is restricted by a difficulty in extracting the
exact sense and context in which vocabulary items are being used by speaker. Indeed, the
final semantic and conceptual representations formed in individuals with WS may well be
shallower, with less abstract information and more perceptually-based detail, suggested by
work examining conceptual knowledge in WS (Johnson & Carey, 1998), as well as by
work on the development of semantic categories and metaphor comprehension (Thomas,
van Duuren, Ansari, Parmigiani, & Karmiloff-Smith, 2002). However, language
development in WS is bolstered by a relative strength in phonological processing that
permits the encoding and production of a range of words. This form of compensation is
evidenced indirectly by basic measures of nonword repetition (Mervis et al., 1999), but
also directly by analyses suggesting that the contribution of phonology to word learning
compared to that of the existing lexicon may be greater in WS than is typical after 5 years
of age (Grant, Karmiloff-Smith, Gathercole, Paterson, Howlin, Davies & Udwin, 1997),
and from word learning tasks suggesting preferential use of phonological over semantic
information (Laing, Hulme, Grant, & Karmiloff-Smith, 2001). The outcome in WS is a
language system that is preferentially geared towards engaging and maintaining social
interaction, and that uses well-formed socially effective vocabulary and phrases with only
approximate semantic underpinnings. Unusual vocabulary in WS is successful in engaging
interest, but closer inspection suggests that it is not supported by appropriate contextual
nuances of meaning (Rossen et al., 1996).
In our view, this sketch emphasises paradigmatic characteristics of the way in which to
explore the origin of cognitive deficits in developmental disorders: explore developmental
Karmiloff-Smith & Thomas, page 36
precursors in infancy, build these into a dynamic account which incorporates both the
social environment of the individual and the capacity for compensation, given the altered
constraints acting on development. Against this type of account, we have contrasted an
approach to language in WS that only references a static model of the normal system and
claims a dissociation between ‘preserved’ syntactic knowledge and ‘impaired’ wordspecific knowledge (Clashen & Almazan, 1998). In fact, we believe that currently there is
no strong support for a grammar-lexicon dissociation in WS, nor indeed a theoretical
consensus amongst linguists or psychologists that this type of dissociation would be
developmentally plausible, given the dynamic and constant interactions between these two
aspects of language during normal acquisition. Thus, while processing in the WS lexicon
appears inefficient, so too does the processing of syntax in WS (as evidenced by
exaggerated patterns of difficulty in WS in sentence processing tasks; see Grant, Valian, &
Karmiloff-Smith, 2002; Zukowski, 2001; and Thomas & Karmiloff-Smith, in press a,
Thomas et al., 2001, for discussion). The most salient and agreed upon dissociation within
the WS cognitive system remains the one first identified by Bellugi and colleagues, of an
imbalance between some aspects of language (such as vocabulary) and visuo-spatial
processing (see WS Cognitive Profile: Mervis, et al., 1999). And even here, evidence of
impairments in spatial vocabulary and spatial representations in WS (Jarrold et al., 2001)
suggests that this dissociation is not static but has implications when spatial and language
systems interact across developmental time.
F. Computational approaches
If links are to be made between genotype and phenotype in developmental disorders,
we have argued that researchers must identify atypical neurocomputational constraints
on plasticity that shape from infancy onwards subsequent trajectories of cognitive
Karmiloff-Smith & Thomas, page 37
development. The preceding two sections have presented psychological evidence
pointing towards the types of constraints that may be atypical in WS language
development. However, notions like phonology and semantics are some way from the
neurocomputational characteristics that may have been altered by early
neurobiological events and subsequent developmental cascades from embryogenesis
onwards. In order to build links between the psychological cognitive level of
description and the neural level, it is essential, in our view, to explore the intermediate
level of computational learning systems (Karmiloff-Smith, 1992). Such learning
systems permit us to build models that encode information at the conceptual level
(e.g., word forms and meanings) – thereby making contact with the psychological
data – but also for those classes of models that share neurocomputational principles
with the underlying substrate, to make contact with the neural level as well.
Connectionist networks are an example of this type of modelling and particularly well
suited to exploring developmental questions (Elman et al., 1996; Karmiloff-Smith,
1992). We would argue that without computational modelling to link the cognitive
and neural levels, researchers are left without a concrete idea of what a
‘developmental process’ might constitute, and what implications it might have in
producing or recovering from deficits. It is one thing to stress the importance of
development in understanding developmental disorders. It is another to begin to
address precisely what the developmental process involves.
Connectionist models have been applied to several developmental disorders, including
WS, SLI, autism, developmental dyslexia, and schizophrenia. In these cases, the
researcher starts with a model formulated to capture the normal trajectory of
development in a given cognitive domain. Over the last fifteen years, connectionist
Karmiloff-Smith & Thomas, page 38
models have been applied to many phenomena within normal cognitive development,
including those in infancy (e.g., categorisation, object-directed behaviour, memory),
in childhood (e.g., Piagetian reasoning tasks such as the balance scale problem,
seriation, and conservation), and in language development (e.g., categorisation of
speech sounds, segmentation of the speech stream into words, vocabulary
development, inflectional morphology, syntax, metaphor, reading). In these models,
it is possible to identify the computational constraints that guide the normal trajectory
of development, such as the architecture of the connectionist network, the activation
dynamics of the system, its input and output representations and its learning rule
(Karmiloff-Smith & Thomas, in press; Thomas & Karmiloff-Smith, 2002, in press, a,
b, c). Either psychological or neuroscientific data are then used to motivate
alterations to these initial constraints in an attempt to capture the atypical trajectory of
development observed in a particular disorder, and any endstate deficits in the adult.
To the intermediate level of computational modelling, one may import downwards
constraints from the psychological level, such as the differences found in
phonological and semantic representations of individuals with in WS (Thomas &
Karmiloff-Smith, in press-a), or in the phonological representations of individuals
with SLI or developmental dyslexia (e.g., Harm & Seidenberg, 1999; Hoeffner &
McClelland, 1993). Or one may import upwards constraints from the neural level,
such as the differences in neuronal density observed in different areas of the brains of
individuals with autism (Cohen, 1998), or differences in the level of dopamine in the
frontal lobes of individuals with schizophrenia (Cohen & Servan-Schreiber, 1992).
Karmiloff-Smith & Thomas, page 39
One of the advantages of computational models is that they allow candidate
explanations of the developmental process to be explicitly formulated, and for the
process to be examined under a range of controlled conditions beyond the scope of
current empirical methods. For example, in one set of simulations (Thomas &
Karmiloff-Smith, in press-b), we compared the effect of the same type of damage
applied to a developmental system either prior to training (to represent the case of a
developmental disorder) or following training (to represent the case of an adult
acquired disorder). This comparison allowed us to assess the contribution of the
developmental process (here a learning rule driving the acquisition of a representative
cognitive domain) in producing patterns of deficits in the final trained state following
various different types of damage. The results revealed a complicated relationship
between patterns of deficits following startstate damage and patterns of deficits
following endstate damage. For certain types of damage (e.g., lesioning), the system
was far more sensitive to damage in the endstate than the startstate – development
served to attenuate the effects of differences in the processing structures. For other
types of damage (e.g., processing noise), the system was far more sensitive to damage
in the startstate than the endstate – development served to exaggerate the effects of
differences in the activation dynamics because in the endstate representations were
already more stable. In some cases, the effects of damage were global (noise,
lesioning), but in other cases the deficit was selective to particular aspects of the
cognitive domain only following damage to the endstate. To the extent these networks
serve as valid models of development, the results emphasise the importance of the
developmental process in determining the patterns of endstate deficits following
alteration to different initial neurocomputational constraints.
Karmiloff-Smith & Thomas, page 40
In another set of simulations (Thomas & Karmiloff-Smith, in press-b), we examined a
developmental system which produced emergent specialisation of particular functions
to particular structures during development. This model allowed us to explore the
conditions under which static explanations of developmental disorders might hold.
That is, if modules are an outcome of development, and one part of the system is
damaged prior to training, does the rest of the system nevertheless develop normally?
Such a condition (which we term residual normality) must hold if strong analogies are
to be drawn between selective cognitive deficits in developmental disorders and those
found in cases of adult damage to previously normally developed brains. Our
simulations revealed that the conditions under which residual normality would hold
are fairly narrow, and in many cases developmentally implausible. A developmental
system suffering initial, selective damage tends to use its remaining resources to
compensate for the initial damage across development. Our computational work
therefore supports the conclusions of Section VI B, where we established that in many
cases, focal brain damage in young healthy children is followed by recovery.
Computational models can also be applied to particular disorders and specific sets of
empirical data. This type of work allows us to evaluate whether particular theoretical
claims are actually sufficient to explain the behavioural deficits observed in a given
developmental disorder. Thus in separate work (Thomas & Karmiloff-Smith, in pressa), we have explored how a connectionist model of normal development of
inflectional morphology (the English past tense) may be applied to the case of WS.
This type of a model attempts to maximise the psychological plausibility of its
constraints, and fit actual patterns of empirical data, from typically developing
children when the constraints are normal and from the target disorder when the
Karmiloff-Smith & Thomas, page 41
constraints have been altered from the outset in line with the available empirical
evidence.
In WS, the evidence on the acquisition of past tense formation initially indicated
difficulty with producing irregular inflections (e.g., think-thought, give-gave)
(Clahsen & Almazan, 1998). However, these preliminary findings have been hard to
replicate, and larger studies have suggested that there exist both problems generalising
inflectional regularities to novel word forms (wug-wugged), and a differential
influence of lexical semantics on inflection in WS (typically developing children
found irregular verbs with more abstract meanings harder to inflect than those with
more concrete meanings, but children and adults with WS performed equally on both)
(Thomas et al., 2001).
In using a computational model to explore this aspect of language acquisition in WS,
we pursued the hypotheses discussed in the previous two sections. We explored
whether manipulations to the initial phonological and semantic representations within
the normal model were sufficient to shift its developmental performance from that of
typically developing children to that of our WS cohort. The results showed that
insufficiently abstract phonological representations (i.e., with reduced redundancy and
between phoneme similarity) were able to capture the problems in inflecting novel
forms, and that weakened semantic representations were able to capture poor
development of irregular inflection, as well as empirical evidence for differential
influence of semantic variables such as abstractness of verb meaning. Importantly,
this model demonstrated that to capture the full range of individual variation in the
empirical data for individuals with WS, multiple initial neurocomputational
Karmiloff-Smith & Thomas, page 42
constraints had to be altered – but that when more than one constraint was altered, the
effects on the subsequent developmental trajectory tended to be interactive rather than
additive. For example, weaker semantic constraints tended to exaggerate the effects of
changes to the phonological constraints.
Although models such as these necessarily contain simplifications (the hallmark of
any modelling process), we believe that they represent one of the vital ways forward
in evaluating developmental deficits with a very concrete notion of the developmental
process in mind. Thus, while our model demonstrated deficits at the end of its
development that matched patterns found in adults with WS, crucially those deficits
were the outcome of differential initial computational constraints and the subsequent
process of development, and did not correspond to the simple removal of any preexisting static structures.
VII.
Conclusion: Contribution to Developmental Theory of the Study of
Language in WS - a Natural Experiment of Nature.
Consistent functional structure appears to emerge in the cognitive systems of normal
adults, and there has long been a debate about the origin of this consistency. At one
extreme, there have been theories of pre-wired, innate modular structure, while at the
other there have been theories of equipotentiality and structure derived from
regularities in the environmental input. Recently, developmental cognitive
neuroscience has moved towards a middle way, the idea of emergentism, whereby
initial constraints in computational properties in the brain ‘seed’ specialisation that
emerges as a product of development (Elman, et al., 1996; Karmiloff-Smith, 1992,
1998; see Thomas, in press, for discussion). However, revealing the nature of these
Karmiloff-Smith & Thomas, page 43
constraints is complex, since normal development confounds the consistency of these
constraints across the population with the consistency of the environment to which
most individuals are exposed. It is thus developmental disorders that may provide a
window on these constraints, since they represent a situation where the constraints
start out differently, and where cognitive-level deficits may be the eventual outcome
of development rather than the initial state.
In the case of genetic developmental disorders with uneven cognitive profiles in their
outcome, we have started to begin to explore more specific links between genotype
and phenotype via case study comparisons (Karmiloff-Smith, Grant, Ewing, Carette,
Metcalfe, Donnai, Read & Tassbehji, 2003). Yet, in our view, progress cannot be
made if researchers continue to characterise deficits within static models, based on
behavioural data from older children and adults (Frangiskakis et al., 1996). This is
because the behavioural deficits that arise from genetic mutations must be traced back
to their infant origins and to the cognitive level of account, with development as a
crucial component of the link (Karmiloff-Smith, 1998; Karmiloff-Smith, Scerif &
Ansari, 2003).
In this paper, we have illustrated this argument with reference to one disorder,
Williams syndrome. We have detailed the genotype, as well as the physical
phenotype, and cognitive phenotype. Then, concentrating on the domain of language
development, we have shown how static accounts of deficits need to be replaced by
developmental accounts. To do so, we have explored similarities to, and differences
from, deficits found in cases of child and adult acquired brain damage. A realistic
causal model of the link between genes and behaviour led us to examine a range of
Karmiloff-Smith & Thomas, page 44
atypical precursors to language development in WS, and then to construct a
preliminary developmental account of WS language acquisition. Finally, we stressed
the utility of computational modelling for linking cognitive and neural levels in the
study of developmental disorders, and for evaluating concrete formulisations of the
developmental process.
In conclusion, we believe that we have demonstrated that developmental disorders can
be highly effective experiments of nature, provided the very process of development
itself remains at the heart of the explanation.
Karmiloff-Smith & Thomas, page 45
References
Anderson, V., Northam, E., Hendy, J., & Wrennall, J. (2001). Developmental
Neuropsychology: A Clinical Approach. Hove, Sussex: Psychology Press.
Anderson, J. A., Silverstein, J. W., Ritz, S. A., & Jones, R. S. (1977). Distinctive
features, categorical perception, and probability learning: Some applications of a
neural model. Psychological Review, 84, 413-451.
Atkinson, J., Anker S., Braddick, O., Nokes, L., Mason, A. & Braddick, F. (2001).
Visual and visuospatial development in young children with Williams syndrome.
Developmental Medicine and Child Neurology, 43(5), 330-337.
Arber, S., Barbayannis, F.A., Hanser, H., Schneider, C., Stanyon, C.A., Bernard, O.,
& Caroni, P. (1998). Regulation of actin dynamics through phosphorylation of
cofilin by LIM-kinase. Nature, 393, 805-809.
Baron-Cohen, S. (1998). Modularity in developmental cognitive neuropsychology:
Evidence from autism and Gilles de la Tourette syndrome. In J. A. Burack, R. M.
Hodapp, & E. Zigler (Eds.), Handbook of Mental Retardation and Development
(pp. 334-348). Cambridge University Press, Cambridge, UK.
Bates, E. & Roe, K. (2001). Language development in children with unilateral brain
injury. In C.A. Nelson, and M. Luciana (Eds.), Handbook of developmental
cognitive neuroscience, (pp. 281-307). Cambridge, MA: MIT.
Bearden, C. E., Wang, P. P., & Simon, T. J. (2002). Williams syndrome cognitive
profile also characterizes Velocardiofacial/DiGeorge Syndrome. American Journal
of Medical Genetics (Neuropsychiatric Genetics), 114, 689-692.
Bellugi, U., Wang, P., & Jernigan, T. L. (1994). Williams syndrome: an unusual
neuropsychological profile. In S. Broman and J. Grafman (Eds.), Atypical
Karmiloff-Smith & Thomas, page 46
cognitive deficits in developmental disorders: Implications for brain function (pp
23-56). Hillsdale, NJ: Erlbaum.
Bertrand, J., Mervis, C., Rice, C.E. & Adamson, L. (1993). Development of joint
attention by a toddler with Williams syndrome. Paper presented at the Gatlinberg
Conference on Research and Theory in Mental Retardation and Developmental
Disabilities, Gatlinberg.
Beuren, A.J., Apitz, J.& Harmjanz, D. (1962). Supravalvular aortic stenosis in
association with mental retardation and a certain facial appearance. Circulation
26,1235-1240.
Bishop, D. V. M. (1997). Cognitive neuropsychology and developmental disorders:
Uncomfortable bedfellows. Quarterly Journal of Experimental Psychology, 50A,
899-923
Black, J.A. & Bonham-Carter, R.E. (1963). Association between aortic stenosis and
facies of severe infantile hypercalcemia. Lancet, 2, 745-748.
Broder, K., Reinhardt, E., Ahern, J., Lifton, R., Tamborlane, W., & Pober, B. (1999).
Elevated ambulatory blood pressure in 20 subjects with Williams syndrome.
American Journal of Medical Genetics, 83, 356-360.
Brown, J., Johnson, M.H., Paterson, S., Gilmore, R.O, Gsödl, M, Longhi, E., &
Karmiloff-Smith, A. (2003). Spatial representation and attention in toddlers with
Williams syndrome and Down syndrome. Neuropsychologia, 41(8), 1037-1046.
Cappelletti, M., Butterworth, B., & Kopelman, M. (2001). Spared numerical abilities
in a case of semantic dementia. Neuropsychologia, 39, 1224-1239.
Cherniske, E.M., Sadler, L.S., Schwartz, D., Carpenter, T.O., & Pober, B.R. (1999).
Early puberty in Williams syndrome. Clinical Dysmorphology, 8, 117-121.
Karmiloff-Smith & Thomas, page 47
Cicchetti, D. (2002). The impact of social experience on neurobiological systems:
illustration from a constructivist view of child maltreatment, Cognitive
Development, 17, 3-4, 1407-1428.
Cicchetti, D., & Tucker, D. (1994). Development and self-regulatory structures of the
mind. Development and Psychopathology, 6, 533-549.
Cipolotti, L., Butterworth, B. & Warrington, E. (1995) Selective impairment of the
manipulation of arabic numerals. Cortex, 31, 73-86.
Clahsen, H., & Almazan, M. (1998). Syntax and morphology in Williams syndrome.
Cognition, 68, 167-198.
Clahsen, H. & Almazan, M. (2001). Compounding and inflection in language
impairment: Evidence from Williams Syndrome (and SLI). Lingua, 111, 729-757.
Cohen, I. L. (1998). Neural network analysis of learning in autism. In D. Stein & J.
Ludick (Eds.), Neural networks and psychopathology (pp. 274-315). Cambridge:
Cambridge University Press.
Cohen, J.D. & Servan-Schreiber, D. (1992). Context, cortex and dopamine: A
connectionist approach to behavior and biology in schizophrenia. Psychological
Review, 99, 45-77.
Cohen, Ansari, Rosen & Karmiloff-Smith (March, 2002). Presentation to the
Williams Syndrome Workshop, London.
Coltheart, M. (2002). Assumptions and Methods. In B. Rapp (Ed), The Handbook of
Cognitive Neuropsychology: What Deficits Reveal About the Human Mind,
Psychology Press.
Curran, M.E., Atkinson, D.L., Ewart, A.K., Morris, C.A., Leppert, M.F., & Keating,
M.T. (1993). The elastin gene is disrupted by a translocation associated with
supravalvular aortic stenosis. Cell, 73,159-168.
Karmiloff-Smith & Thomas, page 48
Deruelle, C., Mancini, J., Livet, M.O., Cassé-Perrot, C., de Schonen, S. (1999).
Configural and local face processing in children with Williams syndrome. Brain
and Cognition, 41, 276-298.
Donnai, D. & Karmiloff-Smith, A. (2000) Williams syndrome: From genotype
through to the cognitive phenotype. American Journal of Medical Genetics:
Seminars in Medical Genetics, 97 (2), 164-171.
Einfeld,, S.L., Tonge, B.J., & Florio, T. (1997). Behavioral and emotional
disturbance in individuals with Williams syndrome. American Journal on Mental
Retardation, 102(1), 45-53.
Elman, J. L., Bates, E. A., Johnson, M. H., Karmiloff-Smith, A., Parisi, D., &
Plunkett, K. (1996). Rethinking innateness: A connectionist perspective on
development. Cambridge, MA.:MIT Press.
Ewart, A.K., Morris, C.A, Ensing, G.J., Loker, J., Moore, C., Leppert, M., & Keating,
M. (1993a). A human vascular disorder, supravalvular aortic stenosis, maps to
chromosome 7. Proceedings of the National Academy of Science, 90, 3226-3230.
Ewart, A.K., Morris, C.A., Atkinson, D., Weishan, J., Sternes, K., Spallone, P., Stock,
D.A., Leppert, M., & Keating, M.T. (1993b). Hemizygosity at the elastin locus in a
developmental disorder, Williams Syndrome. Nature Genetics, 5, 11-16.
Franke, U. (1999). Williams-Beuren syndrome: genes and mechanisms. Human
Molecular Genetics, 8, 1947-1954.
Frangiskakis, J.M., Ewart, A.K., Morris, A.C., Mervis, C.B., Bertrand, J., Robinson,
B.F., Klein, B.P., Ensing, G.J., Everett, L.A., Green, .D., Proschel, C., Cutowski,
N.J., Noble, M., Atkinson, D.L., Odelberg, S.J. & Keating, M.T. (1996). LIMkinase1 hemizygosity implicated in impaired visuospatial constructive cognition.
Cell, 86, 59-69.
Karmiloff-Smith & Thomas, page 49
Galaburda, A.M., Wang, P.P., Bellugi, U. & Rossen, M. (1994). Cytoarchitectonic
anomalies in a genetically based disorder: Williams syndrome. Cognitive
Neuroscience and Neuropsychology, NeuroReport, 5, 753-757.
Gosch, A. & Pankau, R. (1997). Personality characteristics and behaviour problems
in individuals of different ages with Williams syndrome. Developmental Medicine
and Child Neurology, 39(8), 527-533.
Grant, J., Karmiloff-Smith, A., Gathercole, S. A., Paterson, S., Howlin, P., Davies,
M., & Udwin, O. (1997). Phonological short-term memory and its relationship to
language in Williams syndrome. Cognitive Neuropsychiatry, 2, 81-99.
Grant, J., Valian, V., & Karmiloff-Smith, A. (2002). A study of relative clauses in
Williams syndrome. Journal of Child Language, 29, 403-416.
Grice, S., Spratling, M.W., Karmiloff-Smith, A., Halit, H., Csibra, G., de Haan, M., &
Johnson, M.H. (2001). Disordered visual processing and oscillatory brain activity
in autism and Williams Syndrome. Neuroreport, 12, 2697-2700.
Hallidie-Smith, K.A., & Karas, S. (1988). Cardiac anomalies in Williams-Beuren
syndrome. Archives of Disease in Childhood, 63, 809-813.
Harm, M. & Seidenberg, M. S. (1999). Phonology, reading acquisition, and dyslexia:
Insights from connectionist models. Psychological Review, 106, 491-528.
Hebb, D. O. (1949). The organization of behavior. New York: Wiley.
Hoeffner, J. H. & McClelland, J. L. (1993). Can a perceptual processing deficit
explain the impairment of inflectional morphology in developmental dysphasia? A
computational investigation. In E.V. Clark (Ed), Proceedings of the 25th Child
language research forum. Stanford University Press.
Howlin, P., Davies, M. & Udwin, O. (1998a). Cognitive functioning in adults with
Williams syndrome. Journal of Child Psychology and Psychiatry, 39, 183-189.
Karmiloff-Smith & Thomas, page 50
Howlin, P., Davies, M., & Udwin, O. (1998b). Syndrome specific characteristics in
Williams syndrome: To what extent do early behavioural patterns persist into adult
life? Journal of Applied Research in Intellectual Disabilities, 11, 207-226.
Huttenlocher, P.R. (2002) Neural Plasticity: The effects of environment on the
development of the cerebral cortex. Cambridge, MA, US: Harvard
University Press.
Jarrold, C., Hartley, S. J., Phillips, C., & Baddeley, A. D. (2000). Word fluency in
Williams syndrome: Evidence for unusual semantic organisation? Cognitive
Neuropsychiatry, 5, 293-319.
Jarrold, C., Phillips, C. E., Baddeley, A. D., Grant, J., & Karmiloff-Smith (July,
2001). Comprehension of spatial and non-spatial language in Williams syndrome.
Paper presented at the Experimental Psychology Society, Manchester Meeting.
Johnson, S. & Carey, S. (1998). Knowledge enrichment and conceptual change in folk
biology: Evidence from Williams syndrome. Cognitive Psychology, 37, 156-184.
Johnson, M. H., & Morton, J. (1991). Biology and Cognitive Development: The case
of face recognition. Oxford: Blackwell.
Jones, W, Bellugi, U., Lai, Z., Chiles, M., Reilly, J., Lincoln, A., & Adolphs, R.
(2000). Hypersociability in Williams syndrome. Journal of Cognitive
Neuroscience, 12: Supplement, 30-46.
Karmiloff-Smith, A. (1992). Beyond modularity: A developmental perspective on
cognitive science. Cambridge, MA: MIT Press/Bradford Books.
Karmiloff-Smith, A. (1997). Crucial differences between developmental cognitive
neuroscience and adult neuropsychology. Developmental Neuropsychology, 13, 4,
513-524.
Karmiloff-Smith & Thomas, page 51
Karmiloff-Smith, A. (1998). Development itself is the key to understanding
developmental disorders. Trends in Cognitive Sciences, 2, 10, 389-398.
Karmiloff-Smith, A., Grant, J., Ewing, S., Carette, M.J., Metcalfe, K., Donnai, D.,
Read, A.P., & Tassbehji, M. (2003). Using case study comparisons to explore
genotype/phenotype correlations. Journal of American Medical Genetics, 40: 136140.
Karmiloff-Smith, A., Scerif, G., Ansari, D. (2003). Double dissociations in
developmental disorders? Theoretically misconceived, empirically dubious.
Cortex, 39, 161-163.
Karmiloff-Smith, A., Scerif, G., & Thomas, M. S. C. (2002). Different approaches to
relating genotype to phenotype in developmental disorders. Developmental
Psychobiology, 40, 311-322.
Karmiloff-Smith, A. & Thomas, M. (in press). Can developmental disorders be used
to bolster claims from Evolutionary Psychology? A neuroconstructivist approach.
In J. Langer, S. Taylor Parker & C. Milbrath (Eds.) Biology and Knowledge
Revisited: From Neurogenesis to Psychogenesis.
Laing, E., Butterworth, G., Ansari, D., Gsödl, M., Longhi, E., Panagiotaki, G.,
Paterson, S., & Karmiloff-Smith, A. (2002). Atypical development of language and
social communication in toddlers with Williams syndrome. Developmental
Science, 5, 233-246.
Laing, E., Hulme, C., Grant, J., & Karmiloff-Smith, A. (2001). Learning to read in
Williams syndrome: Looking beneath the surface of atypical reading development.
Journal of Child Psychology and Psychiatry, 42, 729-739.
Leslie, A.M. (1992). Pretence, autism, and the Theory-of-Mind-Module. Current
Directions in Psychological Science, 1, 18-21.
Karmiloff-Smith & Thomas, page 52
Masataka, N. (2000). Information from speech and gesture is integrated when
meanings of new words are categorized in normal young children but not in
children with Williams syndrome. Bulletin of the Japanese Cognitive Science
Society, 7, 37-51.
Masataka, N. (2001). Why early linguistic milestones are delayed in children with
Williams syndrome: late onset of hand banging as a possible rate-limiting
constraint on the emergence of canonical babbling. Developmental Science, 4,
158-164.
Mervis, C., & Bertrand, J. (1997). Developmental relations between cognition and
language: Evidence from Williams Syndrome. In L. B. Adamson & M. A. Romski
(Eds.), Research on communication and language disorders: Contributions to
theories of language development (pp. 75-106). New York: Brookes.
Mervis, C. B., Morris, C. A., Bertrand, J., & Robinson, B. F. (1999). William
Syndrome: Findings from an integrated program of research. In H. Tager-Flusberg
(Ed), Neurodevelopmental disorders (pp. 65-110). Cambridge, Mass: MIT Press.
Mervis, C.B., Robinson, B.F., Bertrand, J., Morris C.A., Klein-Tasman, B.P. &
Armstrong, S.C. (2000). The Williams syndrome cognitive profile. Brain &
Cognition, 44, 604-628.
Mervis, C.B., Robinson, B.F., Bertrand, J., Morris C.A., Klein-Tasman, B.P., Bonita.,
P., & Armstrong,, S.C. (2002). The Williams Syndrome Cognitive Profile. Brain
and Cognition, 44(3): 604-628
Morris, C.A., Demsey, S.A., Leonard, C.O., Dilts, C., & Blackburn, B.L. (1988).
Natural history of Williams syndrome: Physical characteristics. Journal of
Paediatrics, 113(3), 18-326.
Karmiloff-Smith & Thomas, page 53
Morris, C.A., Thomas, I.T., & Greenberg F. (1993). Williams syndrome: Autosomal
dominant inheritance. American Journal of Medical Genetics, 47, 478-481.
Nazzi, T., & Karmiloff-Smith, A. (2002). Early categorization abilities in young
children with Williams syndrome. NeuroReport, 13(10):1259-1262.
Nazzi, T., Paterson, S. & Karmiloff-Smith, A. (2003). Early word segmentation by
infants and toddlers with Williams syndrome. Infancy, 4 (2), 251-271
Neville, H. J., Mills, D. L., & Bellugi, U. (1994). Effects of altered auditory
sensitivity and age of language acquisition on the development of language–
relevant neural systems: preliminary studies of Williams syndrome. In S. Broman
and J. Grafman (Eds.), Atypical cognitive deficits in developmental disorders:
Implications for brain function (pp. 67-83). Hillsdale, NJ: Erlbaum.
Pankau, R., Gosch, A., Simeoni, E., & Wessel, A. (1993). Williams-Beuren
syndrome in monozygotic twins with variable expression. American Journal of
Medical Genetics, 47, 475-477.
Paterson, S. J. (2000). The development of language and number understanding in
Williams Syndrome and Down's Syndrome: Evidence from the infant and mature
phenotypes. Unpublished doctoral thesis. University College London.
Paterson, S. J., Brown, J. H., Gsödl, M. K., Johnson, M. H., & Karmiloff-Smith, A.
(1999). Cognitive modularity and genetic disorders. Science, 286, 2355-2358.
Patterson, Karalyn E., 1981. Neuropsychological approaches to the study of reading.
British Journal of Psychology, 72.151-174.
Pennington, B.F. (2001). Genetic methods. In C.A. Nelson and M. Luciana (Eds.),
Handbook of developmental cognitive neuroscience (pp.149-158). Cambridge,
MA: MIT Press.
Karmiloff-Smith & Thomas, page 54
Peoples, R., Franke, Y., Wang, Y.-K., Perez-Jurado, L., Paperna, T., Cisco, M., &
Franke, U. (2000). A physical map, including a BAC/PAC clone contig of the
Williams-Beuren Syndrome deletion region at 7qll.23. American Journal of
Human Genetics, 66, 47-67.
Perez-Juralo, L.A., Peoples, R., Kaplan, P., Hamel, B.C.J., & Franke, U. (1996).
Molecular definition of the chromosome 7 deletion in Williams syndrome and
parent-of-origin effects on growth. American Journal of Human Genetics, 59, 781792.
Pinker, S. (1991). Rules of language. Science, 253, 530-535.
Pinker, S. (1994). The language instinct. London: Penguin.
Pinker, S. (1997). How the mind words. New York: Norton.
Pinker, S. (1999). Words and rules. London: Weidenfeld & Nicolson.
Proschel, C., Blouin, M. J., Gutowski, N. J., Ludwig, R., & Noble, M. (1995).
L1MK1 is predominantly expressed in neural tissue and phosphorylates serine,
threonine and tyrosine residues in vitro. Oncogene, 11, 1271 – 1281.
Rae. C., Karmiloff-Smith, A., Lee, M.A., Dixon, R.M., Blamire, A.M., Thompson,
C.H., Grant, J., Styles, P. & Radda, G.K.. (1998) Brain Biochemistry in Williams
Syndrome. Evidence for a role of the cerebellum in cognition? Neurology, 51, 3340.
Rapp, A., & Caramazza, A. (2002) Selective difficulties with spoken nouns and
written verbs: A single case study. Journal of Neurolinguistics, 15, 373-402.
Robinson, W.P., Waslynka, J., Bernasconi ,F., Wang, M., Clark, S., Kotzot, D., &
Schinzel, A. (1996). Delineation of 7q11.2 deletions associated with WilliamsBeuren syndrome and mapping of repetitive sequences to within and to either side
of the common deletion. Genomics, 34,17-23.
Karmiloff-Smith & Thomas, page 55
Rossen, M., Klima, E. S., Bellugi, U., Bihrle, A., & Jones, W. (1996). Interaction
between language and cognition: Evidence from Williams syndrome. In J. H.
Beitchman, N. Cohen, M. Konstantareas & R. Tannock (Eds.), Language Learning
and Behaviour (pp. 367-392). New York, NY: Cambridge University Press.
Rourke, B. P. (1987). Syndrome of non-verbal learning disabilities: The final
common pathway of white matter disease/dysfunction. The Clinical
Neuropsychologist, 1, 209-234.
Rourke, B. P. (1989). Nonverbal learning disabilities. New York: Guilford Press.
Rourke, B. P., & Tsatsanis, K. D. (1996). Syndrome of nonverbal learning
disabilities: Psycholinguistic assets and deficits. Topics in Language Disorders, 16,
30-44.
Scott, P., Mervis, C. B., Bertrand, J., Klein, B. P., Armstrong, S. C., & Ford, A. L.
(1995). Semantic organization and word fluency in 9- and 10-year-old children
with Williams syndrome [special issue]. Genetic Counseling, 6, 172-173.
Shallice, T. (1988). From neuropsychology to mental structure. Cambridge:
Cambridge
University Press.
Singer Harris, N. G., Bellugi, U., Bates, E., Jones, W., & Rossen, M. (1997).
Contrasting profiles of language development in children with Williams and Down
syndromes. Developmental Neuropsychology, 13, 345-370.
Smith, N. (1999). Chomsky: Ideas and ideals. Cambridge: Cambridge University
Press.
Smith, N. & Tsimpli, I-M. (1995). The Mind of a Savant: Language, Learning and
Modularity, Oxford: Blackwell.
Karmiloff-Smith & Thomas, page 56
Stevens, T., & Karmiloff-Smith, A. (1997). Word learning in a special population: do
individuals with Williams syndrome obey lexical constraints? Journal of Child
Language, 24, 737-765.
Tager-Flusberg, H., Boshart, J., & Baron-Cohen, S. (1998). Reading the windows to
the soul: Evidence of domain-specific sparing in Williams syndrome. Journal of
Cognitive Neuroscience, 10(5), 631-639.
Tassabehji, M., Metcalfe, K., Fergusson, W.D., Carette, M.J., Dore, J.K., Donnai, D.,
Read, A.P., Proschel, C., Gutowski, N.J., Mao, X., & Sheer, D. (1996). LIM-kinase
deleted in Williams syndrome. Nature Genetics, 13, 272-273.
Temple, C. (1997). Developmental cognitive neuropsychology. Psychology Press.
Temple, C., Almazan, M., & Sherwood, S. (2002). Lexical skills in Williams
syndrome: a cognitive neuropsychological analysis. Journal of Neurolinguistics,
15(6), 463-495.
Temple, C & Clahsen, H (in press). How connectionist simulations fail to account for
developmental disorders in children. Behavioral and Brain Sciences.
Thomas, M. S. C. (in press). Multiple causality in developmental disorders:
Methodological implications from computational modelling. Developmental
Science.
Thomas, M. S. C., Dockrell, J., Messer, D., Parmigiani, C., Ansari, D., & KarmiloffSmith, A. (2002). Naming in Williams syndrome. Manuscript submitted for
publication.
Thomas, M. S. C., Grant, J., Gsödl, M., Laing, E., Barham, Z., Lakusta, L., Tyler, L.
K., Grice, S., Paterson, S. & Karmiloff-Smith, A. (2001). Past tense formation in
Williams syndrome. Language and Cognitive Processes, 16, 143-176.
Karmiloff-Smith & Thomas, page 57
Thomas, M. S. C. & Karmiloff-Smith, A. (2002). Modelling typical and atypical
cognitive development. In U. Goswami (Ed.), Handbook of Childhood Cognitive
Development (pp. 575-599). Oxford: Blackwell Publishers.
Thomas, M. S. C. & Karmiloff-Smith, A. (in press, a). Modelling language acquisition
in atypical phenotypes. Psychological Review.
Thomas, M. S. C. & Karmiloff-Smith, A. (in press, b). Are developmental disorders like
cases of adult brain damage? Implications from connectionist modelling. Behavioral
and Brain Sciences.
Thomas, M. S. C. & Karmiloff-Smith, A. (2003). Connectionist models of
development, developmental disorders and individual differences. In R. J.
Sternberg, J. Lautrey, & T. Lubart (Eds.), Models of Intelligence:
International Perspectives, (p. 133-150). American Psychological
Association.
Thomas, M. S. C., van Duuren, M., Ansari, D., Parmigiani, C., & Karmiloff-Smith, A.
(2002). The development of semantic categories and metaphor comprehension in
Williams syndrome. Manuscript in preparation.
Udwin, O. & Yule, W. (1991). A cognitive and behavioural phenotype in Williams
syndrome. Journal of Clinical and Experimental Neuropsychology, 13, 232-244.
Urban. Z., Helms, C., Fekete, G., Csiszar, K., Bonnet, D., Munnich, A., Donis-Keller,
H., & Boyd, C.D. (1996). 7q11.23 deletions in Williams syndrome arise as a
consequence of unequal meiotic crossover. American Journal of Human Genetics,
59, 958-962.
Volterra, V., Capirci, O., & Caselli, M. C. (2001). What atypical populations can
reveal about language development: The contrast between deafness and Williams
syndrome. Language and Cognitive Processes, 16, 219-239.
Karmiloff-Smith & Thomas, page 58
Wexler K (1996). The development of inflection in a biologically based
Theory of language acquisition. In Toward a Genetics of Language, ML Rice
(Ed). Mahwah, NJ: Lawrence Erlbaum Associates, Inc.
Williams, J.C.P., Barratt-Boyes, B.G. & Lowe, J.B. (1961). Supravalvular aortic
stenosis. Circulation, 24, 1311-1318.
Winter, M., Pankau, R., Amm, M., Gosch, A,, & Wessel, A. (1996). The spectrum of
ocular features in the Williams-Beuren syndrome. Clinical Genetics, 49, 28-31.
Zukowski, A. (2001). Uncovering grammatical competence in children with Williams
syndrome. Unpublished doctoral thesis, Boston University.
1
We thank our colleague, Julia Grant, for drawing our attention to the fact that our original + sign
implied an additive notion whereas we had in mind a multiplicative notion. In fact, more precisely the
equation should read f(x, y, z) where x is the initial existing neurocomputational state, y defines the
way it changes in response to activity, and z is the internal/external environment. But the equation used
above is more self explanatory.
2
Whether the subsystem falls inside or outside the normal range of function is a dichotomy enforced
by measurement on a standardised test. Further, more subtle measures are required to establish the
degree to which functional structure actually falls within the normal range of variation.